Modeling the Interaction between Culture and Institutional ... Modeling the Interaction between

  • View
    0

  • Download
    0

Embed Size (px)

Text of Modeling the Interaction between Culture and Institutional ... Modeling the Interaction between

  • Modeling the Interaction between Culture and Institutional Performance, with Implications for Institutional Sequencing and

    Robust Design

    Jenna Bednar & Scott Page University of Michigan

  • Research Motivation

    • Laws are embedded in context. They are subject to interpretation; that interpretation is influenced by other institutions as well as culture.

  • Research Motivation

    • Laws are embedded in context. They are subject to interpretation; that interpretation is influenced by other institutions as well as culture.

    • Great, that’s intuitive. But now what?

  • We can measure cultural differences…

  • Source: data courtesy Ron Inglehart.

  • Cultural Differences have Meaningful Effects on Institutional Performance

  • Cultural Differences Affect Cooperation

  • Evidence of lingering effects of culture and pre-existing institutions, but no means to disentangle the two.

    Legacy: “A new democracy can inherit a constitution, a number of laws and regulations, a large army, or an inefficient bureaucracy from the previous regime”

  • Disagreement about policy prescriptions

  • Research Motivation

    • Laws are embedded in context. They are subject to interpretation; that interpretation is influenced by other institutions as well as culture.

    • Great, that’s intuitive. • We need a model of how culture and institutions

    affect one another.

  • Larger Project

    Culture Institutions

  • Larger Project

    Bednar and Page (2007) “Can Game(s) Theory Explain Culture? The Emergence of Cultural Behavior Within Multiple Game” Rationality and Society Bednar, Chen, Liu, and Page (2012) “Behavioral Spillovers and Cognitive Load in Multiple Games: An Experimental Study.” Games and Economic Behavior.

    Behavioral spillover (culture)

    Culture Institutions

  • Game(s) Theory

    People interact in multiple strategic settings and their behavior in one setting can spill over into other settings. These patterns of behavior shape culture.

  • 3 Sets of Results

    • Spillovers (using experiments) • Sequencing (using mathematical model) • Diversity (using ABM)

  • Results 1: Behavioral spillovers affect Institutional Performance

  • Our Games

    7,7 2,10

    10,2 4,4

    7,7 2,9

    9,2 10,10

    7,7 4,14

    14,4 5,5

    7,7 4,11

    11,4 5,5

    Self Interest Prisoner’s Dilemma

    Strong Alternation Weak Alternation

  • Variation in Experiments

    Control Trtmt 1 Trtmt 2 Trtmt 3 PD: % CC 56 42 41 40 SA: % ALT 71 48 48 38 WA: % ALT 36 21 18 37

  • Why?

  • Standard Experiment

    7,7 2,10

    10,2 4,4 D

  • Our Experiments

    7,7 2,10

    10,2 4,4

    7,7 4,14

    14,4 5,5

  • Variation in Experiments

    Control w/ Self Interest w/ SA* or

    PD w/ WA or

    SA* PD: % CC 56 42 41* 40 SA: % ALT 71 48 48 38 WA: % ALT 36 21 18 37*

  • Two Theories for Observed Variation

    • Cognitive Overload • Behavioral Spillovers

  • Two Theories

    • Cognitive Overload – Rationale: Multitasking leads to confusion and

    suboptimal decision-making

  • Two Theories

    • Cognitive Overload • Behavioral Spillovers

    – Rationale: in solving new problem, apply behavior from existing repertoire (ie Agents apply same strategies in different contexts)

  • Cultures are coherent; we have behavioral expectations based on culture

  • Research questions

    • When will behavioral spillovers exist? • Can we predict their direction? • We need a measure of game difficulty to allow

    us to compare games

  • ENTROPY

    • Measure of amount of info needed to describe the distribution of outcomes

    • Measure variation in control responses • Correlated with cognitive load • In 2X2, bounded between [0,2], where H = 0 is

    all observations in one cell and H=2 uniform across all four cells

  • ENTROPY

    • aka, Measure of surprise • Low entropy: you know what people will do • High entropy: their reactions may surprise you

  • Hypotheses

    • Low entropy games will not be subject to influence

    • Behavioral spillovers will run from low entropy to high or between highs

    • Cognitive Load highest in high entropy games, so least behavioral pull

  • Ensemble Effect on SI play

    • Easiest game: NO EFFECT. – Behavior across rounds > 99% selfish in controls

    and ensembles.

  • SI Effect on other Games

    • Predicted: More SS in PD, WA, and SA when each is paired with SI than in control

  • SI’s Effect on Behavior in other Games

    • Predicted: More SS in PD, WA, and SA when each is paired with SI than in control

    • Significant Differences for PD and SA: – PD vs PD+SI: 0.0780 / 0.1049 – SA vs SA+SI: 0.1220 / 0.0800 – WA vs WA + SI: 0.3966 / 0.5787 (p-values, whole series / last 100 rounds)

  • Behavioral Spillovers: PD Games

    • (PD+SA) : (PD+SI) – More ALT w/SA : 0.0455 / 0.0452 – More SS w/SI : 0.0736 / 0.1049

    • (PD+SA) : (PD+WA) – More ALT w/SA : 0.0910 / 0.0727 – More SS w/WA : 0.1063 / 0.0754 (p-values, whole series/last 100 rounds)

  • What our work means for culture and institutional

    performance

  •  

    Will personal exchange scale, and if not, why not?

  • + +

  • + +

  • Spillovers: Key Insights

    • Institutions may not perform as expected due to the presence of behavioral spillovers.

    • Institutions with multiple “reasonable” reactions are most susceptible.

    • Response to susceptible institutions is driven by behavior in easier games (w dominant strategy)

  • Results 2: Sequencing of Institutions affects

    Institutional Performance

  • The Model

    • Games arrive in sequence

    • Initial Behavior: – Efficient equilibrium strategy (1-γ) – Equilibrium strategy of nearest game (γ)

    • Best response dynamics (Nash 1951)

  • 16-Θ,16-Θ 4,4

    4,4 Θ,Θ

    Tradition Innovate

    Tradition Innovate

  • T I

    0 8 16

    Efficient Efficient θ

  • 0 6 10 16

    Trad Immune Susceptible Innov Immune

    Proof

  • 0 8 16

    ? Θ

    Game 1: Θ = 9

  • 0 8 16

    I Θ

    Innovate.

  • 0 8 16

    ? Θ

    I Θ

  • I Θ

    I Θ

    0 8 16

  • Sequencing: Key Insights

    • Path dependence widespread with moderate spillovers

    • Larger behavioral spillover increases susceptible region and path dependence.

    • Efficient paths require clearer incentives early (1- dim)

    • Efficient paths maintain path dependence • Endogenous Institutional change occurs too late • Efficient paths require weak punishment

    (General)

  • Results 3: Behavioral Diversity affects Institutional Performance

  • Effective Number of Behaviors: 2.3

    Bottom Left

  • Effective Number of Behaviors: 4.58

    PD

  • Diversity: Key Insights

    • Diverse behaviors increase adaptability, resilience

  • SI

    SH

    PD

    KE ALT

    BL

    TR

    TR

    SI

    SH

    PD

    KE ALT

    BL

    TR

    SI SI

    SH

    PD

    KE ALT

    BL

    TR

    SH

    SI

    SH

    PD

    KE ALT

    BL

    TR

    PD

    SI

    SH

    PD

    KE ALT

    BL

    TR

    BL

    SI

    SH

    PD

    KE ALT

    BL

    TR

    AL

    SI

    SH

    PD

    KE ALT

    BL

    TR

    KE

    X

  • SI

    SH

    PD

    KE ALT

    BL

    TR

    TR

    SI

    SH

    PD

    KE ALT

    BL

    TR

    SI SI

    SH

    PD

    KE ALT

    BL

    TR

    SH

    SI

    SH

    PD

    KE ALT